xref: /aosp_15_r20/external/ComputeLibrary/src/gpu/cl/operators/ClTransposedConvolution.cpp (revision c217d954acce2dbc11938adb493fc0abd69584f3)
1 /*
2  * Copyright (c) 2022 Arm Limited.
3  *
4  * SPDX-License-Identifier: MIT
5  *
6  * Permission is hereby granted, free of charge, to any person obtaining a copy
7  * of this software and associated documentation files (the "Software"), to
8  * deal in the Software without restriction, including without limitation the
9  * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
10  * sell copies of the Software, and to permit persons to whom the Software is
11  * furnished to do so, subject to the following conditions:
12  *
13  * The above copyright notice and this permission notice shall be included in all
14  * copies or substantial portions of the Software.
15  *
16  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
19  * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
22  * SOFTWARE.
23  */
24 #include "src/gpu/cl/operators/ClTransposedConvolution.h"
25 
26 #include "arm_compute/core/Error.h"
27 #include "arm_compute/runtime/CL/CLScheduler.h"
28 #include "src/common/utils/Log.h"
29 #include "src/gpu/cl/kernels/ClTransposedConvolutionKernel.h"
30 
31 namespace arm_compute
32 {
33 namespace opencl
34 {
configure(const CLCompileContext & compile_context,const ITensorInfo * input,const ITensorInfo * weights,const ITensorInfo * biases,ITensorInfo * output,const PadStrideInfo & deconv_info)35 void ClTransposedConvolution::configure(const CLCompileContext &compile_context, const ITensorInfo *input, const ITensorInfo *weights,
36                                         const ITensorInfo *biases, ITensorInfo *output, const PadStrideInfo &deconv_info)
37 {
38     ARM_COMPUTE_ERROR_ON_NULLPTR(input);
39     ARM_COMPUTE_LOG_PARAMS(input, weights, biases, output, deconv_info);
40     auto kernel_object = std::make_unique<kernels::ClTransposedConvolutionKernel>();
41     kernel_object->set_target(CLScheduler::get().target());
42     kernel_object->configure(compile_context, input, weights, biases, output, deconv_info);
43     _transposed_conv_kernel = std::move(kernel_object);
44 }
45 
validate(const ITensorInfo * input,const ITensorInfo * weights,const ITensorInfo * biases,const ITensorInfo * output,const PadStrideInfo & deconv_info)46 Status ClTransposedConvolution::validate(const ITensorInfo *input, const ITensorInfo *weights, const ITensorInfo *biases,
47                                          const ITensorInfo *output, const PadStrideInfo &deconv_info)
48 {
49     ARM_COMPUTE_RETURN_ON_ERROR(kernels::ClTransposedConvolutionKernel::validate(input, weights, biases, output, deconv_info));
50     return Status{};
51 }
52 
run(ITensorPack & tensors)53 void ClTransposedConvolution::run(ITensorPack &tensors)
54 {
55     CLScheduler::get().enqueue_op(*_transposed_conv_kernel.get(), tensors, false);
56 }
57 } // namespace opencl
58 } // namespace arm_compute
59